Models And Methods Of Extracting Forest Structure Information By Polarimetric SAR Interferometry

Forest plays an important role as a natural resource in global hydrology, ecology, climate, carbon (biomass) storage and carbon dynamic cycles. The main parameters in survey of forestry resources are forest categories, forest structure, height, biomass, and so on. With the development of remote sensing technology, polarimetric SAR interferometry (POLInSAR) technology which is based on the coherent combination of radar interferometry and polarimetry, has become an irreplaceable technology for survey of forestry resources because of its unique all-weather and low-cost advantages. In forest mapping, particularly quantitative retrieval of forest parameters, it plays an increasing important role. The researches on forest classification, retrieval model of forest height and biomass estimation model using POLInSAR are essential in the studies of SAR remote sensing of forest and are of great significance for application in forest. On the basis of understanding of POLInSAR theory, models and method for extracting forest structure information by POLInSAR are studied in detail in the paper. The main results are as follows:1. Based on the complementary information contained in polarimetric and interferometric SAR data, unsupervised classification approach of forest is studied in detail by using fuzzy clustering and polarimetric interferometric coherence optimization technique. The proposed method employs scattering mechanisms to indentify forest area from polarimetric SAR data. Then the forest area is further segmented by parameters A1 and A2 obtained by polarimetric interferometric coherence optimization algorithm. A robust unsupervised fuzzy C means (FCM) classifier initialized with the results of the segmentation is applied to the polarimetric interferometric coherency data sets corresponding to the forest area. As the initial categories are defined by the characteristics of the data itself, this not only solves the problem that the initial value of FCM algorithm is difficult to identify, but also avoid the fact that the A1/A2 zone boundaries were determined in somewhat arbitrary ways. So the proposed method has good performance.2. An inversion method of forest height with good performance is proposed. Several available forest height inversion models such as random volume (RV), digital elevation model (DEM), random volume over ground (RVoG), and so on are studied in detail. Then this paper investigates that to what extent do interferometric coherence optimization in radar polarimetry and non-volumetric scattering decorrelation improve the performance of forest height inversion methods. The effects of the extinction coefficient and ground-to-volume scattering ratio in RVoG model on polarimetric interferometric coherence are analyzed by means of numerical simulation. On this basis, an integrated inversion method, which combines coherence phase with coherence amplitude information and includes polarization coherence optimization, ground-to-volume scattering ratio and compensation of non-volume scattering decorrelation, is proposed and discussed. The results show that the method is robust and accurate.3. Based on polarization coherence tomography (PCT) technique, the factors possibly affecting the radar relative reflectivity function are investigated in detail and the result is found out that the characteristic parameters extracted from the average relative reflectance functions are sensitive to the biomass. The effects of the extinction coefficient and ground-to-volume scattering ratio in RVoG model on relative reflectivity function are analyzed by means of numerical simulation. Then by applying PCT to L-band POLInSAR simulations of forest scene, the effects of the polarization, forest type and density on relative reflectivity function through extinction coefficient and ground-to-volume scattering ratio are discussed. Furthermore, based on repeated pass DLR E-SAR L-band airborne POLInSAR data, relative reflectivity function curve of different levels of typical stand AGB are analyzed. Then, it is concluded that the shape features of the relative reflectivity function curve are closely related to forest AGB.4. A forest AGB estimation model with good accuracy is constructed. Based on forest stand average relative reflectivity function curve, the nine characteristic parameters are defined and used to construct forest AGB estimation model by multiple linear stepwise regression analysis method. The model is evaluated and the forest AGB estimation accuracy is good because of the stand vertical structure information considered comprehensively in the AGB estimation model.